Annotation-Free Whole-Slide Image Analysis Method to Assess Immune Infiltration in Colorectal Cancer

Area of Science:

• Computational pathology
• Immunohistochemistry
• Colorectal cancer research

Background:

• Tumor-infiltrating lymphocytes (TILs) are critical in antitumor responses and patient outcomes.
• Accurate quantification of TILs is essential for prognostic assessment in colorectal cancer.

Purpose of the Study:

• To develop an AI pipeline for automated TIL quantification in H&E images without nuclei annotation.
• To create a tumor stroma immune (TSI) score for predicting survival in colorectal cancer patients.

Main Methods:

• An AI pipeline was created to quantify TILs in H&E images using IHC-guided labels.
• The study included development (n=557) and validation (n=439) cohorts of colorectal cancer patients.
• The TSI score was calculated based on immune infiltration in tumor stroma.

Main Results:

• Strong correlation observed between TILs in consecutive H&E and IHC stained sections.
• The TSI score independently predicted prognosis, with higher scores indicating better survival (HR 0.54, P=.001; HR 0.68, P=.031).
• The TSI score improved model predictive accuracy (C-index 0.700 vs 0.679; 0.689 vs 0.677).

Conclusions:

• IHC images provide valuable TIL density references for corresponding H&E images.
• The TSI score demonstrates significant potential for predicting overall survival in colorectal cancer.